A Multi-object Tracking Method Based on Bounding Box and Features

Multiple object tracking using feature fusion in hierarchical LSTMs

“…The accuracy of the tracking‐by‐detection approach for multiple object detection primarily depends on computations of the pair‐wise distance between detections in different time stamps. In this context, the earlier works on object tracking were primarily based on appearance‐based matching, or based on bounding box overlaps [3, 10]. The seminal work by Cominicu et al [11] presented the idea of detection matching in higher dimension space using kernels.…”